C57BL/6 wild-type (WT) or C57BL/6.IFNγ deficient (IFNγ^-/-) mice were bred and housed at the i3S Animal House Facility, and infected under ABSL3 conditions. Both male and female mice were used, ranging between 8 and 12 weeks of age at the time of infection. Animals were kept under a controlled temperature (20–24°C), humidity (45–65%), and light cycle (12h light/dark) in specific-pathogen free conditions. Water and food were provided ad libitum. Aerosol infection was performed using an inhalation exposure system (Glas-Col) (28, 29). Infection level was determined based on the lung bacterial burden 3 days post-infection: <200 CFUs corresponded to a low dose, while >500 CFUs was considered a high dose. Infected mice were weighed every week or every 2 days upon showing signs of disease. Weight loss of over 20% or lack of responsiveness to physical stimulation were considered humane endpoints, at which point the experiments were terminated. Mice were euthanized by CO₂ inhalation, during which efforts were made to minimize suffering.

Organs were aseptically excised and processed as described previously (28, 29). Lungs were digested using Collagenase type IV (Gibco, Cat. #17104019) followed by physical disruption and filtering in a 70 µm mesh. Blood was collected via cardiac puncture. BM was collected by flushing the tibias and femurs with PBS + 2% FBS (Gibco, Cat. #10500064). Single-cell suspensions were used for bacterial burden determination, flow cytometry, and RNA analysis.

M. tuberculosis HN878 isolate was grown and stored as described previously (28, 29). Viable bacteria were determined by serial dilution and colony forming unit (CFU) enumeration after 21–28 days of incubation at 37°C in 7H11 agar plates (Cat. #212203, BD Biosciences). Bacterial quantification in infected lungs and BM was performed upon organ homogenization using the same procedure.

Mouse blood, BM, and lung cell suspensions were stained for surface antigens (30 min; 4°C) and fixed for 20 min in 4% paraformaldehyde-PBS after erythrocyte lysis. For the analysis of BM precursors, the mature lineage (Lin⁺) was depleted using an EasySep™ Mouse Streptavidin RapidSpheres™ Isolation Kit (Cat. #19860, STEMCELL Technologies) prior to Lin^- staining with specific antibodies for precursor and progenitor markers. Dead cells were excluded from the analyses using ZombieAqua (Cat. #423102, Biolegend) or ZombieGreen (Cat. #423112, Biolegend) viability dyes. Cells were acquired on a BD FACS Canto II or BD Fortessa II. Data were analyzed using FlowJo (version 10.1.r7). Dimensionality reduction of concatenated data was performed with the Uniform manifold approximation and projection (UMAP) FlowJo plug-in (version 3.1) (30). All antibodies used are listed in Supplementary Table 1 . Gating strategies are shown in Supplementary Figures 1A, B and Supplementary Figure 2A .

Total RNA was extracted from mouse lung or BM cell suspensions using TripleXtractor (Cat. #GB23.0100, GRiSP) and converted to cDNA using a ProtoScript^® II First Strand cDNA Synthesis Kit (Cat. #E6560L, NE Biolabs). Relative CSF2 gene expression quantification was performed using quantitative PCR and Taqman probe assay (Cat. #4331182; Assay ID #Mm01290062_m1, Thermo Fisher Scientific). Normalization of target gene abundance was performed using Hprt as the reference gene (Cat. #4331182; Assay ID #Mm03024075_m1, Thermo Fisher Scientific). Gene expression quantification for Csf1, Csf3, Tnf, Il1b, Il6, Ifng, and Ifnb was performed using quantitative PCR with SYBR green (Cat. #1725120, Bio-Rad) and intron spanning oligonucleotides specific for the target genes. Normalization of target gene abundance was performed using Ubiquitin as the reference gene. Melting and standard curves and RQ values were determined for each gene. Sequences of the oligonucleotides used are in Supplementary Table 2 .

Mouse lungs were fixed in 10% buffered formalin and embedded in paraffin. Serial 3-µm sections were used for Hematoxylin and Eosin (H&E) or immunofluorescence staining and analysis. Morphometric analysis of lung pathology areas in H&E-stained histological sections was performed using the software Interactive Learning and Segmentation Toolkit (Ilastik version 1.3.3) and CellProfiler Analyst (version 3.1.5). Probability maps for the whole lung and lesion area were created in Ilastik and calculated and analyzed using CellProfiler Analyst software (28). Pathological scoring analysis of H&E-stained histological sections was performed in a blind way, and the histopathological features were scored according to the parameters in Supplementary Table 3 .

Paraffin-embedded lung sections were stained for myeloperoxidase (Mpo) and nitric oxide synthase (Nos2), as described previously (28, 29, 31). Antigen recovery was performed by incubating slides at 96°C in HIER Buffer L (Cat. #TA135HBL, Epredia) for 45 min. Slides were then incubated overnight with goat anti-mouse Mpo (Cat. #AF3667, R&D systems; 1:40) and rabbit anti-mouse Nos2 (M-19, Cat. #sc-650, Santa Cruz Biotechnology, 1:50) at 4°C followed by incubation with secondary antibodies donkey anti-goat IgG Alexa Fluor 488 conjugated (Cat. #A-11055, Invitrogen) and donkey anti-rabbit IgG Alexa Fluor 647 conjugated (Cat. #A-31573, Invitrogen) for 2 h at room temperature. Slides were stained with DAPI (Cat. #422801, Biolegend; 1:1000 dilution), and glass coverslips were mounted using Fluoroshield (Cat. #F6182, Sigma-Aldrich). Images were acquired using IN Cell Analyzer 2000 (GE Healthcare) and analyzed using IN Cell Developer software (version 1.9) and ImageJ (version 1.53t) (32, 33).

BM-derived macrophages (BMDMs) were generated from BM cellular suspensions as previously described (29, 34). Cells were plated in Petri dishes and differentiated into macrophages using DMEM (Cat. #10566016, Gibco), completed with 1% sodium pyruvate (Cat. #S8636, Sigma-Aldrich), 1% HEPES (Cat. #15630106, Gibco), and 10% FBS (Cat. #10500064, Gibco) and supplemented with 20% L-cell conditioned medium (V/V). Once differentiation was complete (7 days), BMDMs were transferred to cell culture plates. BM monocytes were purified from BM cellular suspensions using an EasySep™ Mouse Monocyte Isolation Kit (Cat. #19861, STEMCELL Technologies), following the manufacturer’s protocol, and plated in cell culture plates with RPMI medium (Cat. #61870036, Gibco), completed as described above for DMEM. BMDMs were plated in 96-well plates (100,000 cells per well) for CFU determination or in 24-well plates (500,000 cells per well) for supernatant collection. Purified monocytes (purity >95%) were plated in 96-well plates (100,000 cells per well).

Peripheral blood was collected and processed to obtain PBMCs as previously described (34). CD14⁺ monocytes were magnetically isolated using CD14 MicroBeads (Cat. #130-050-201, Miltenyi Biotec), following the manufacturer’s protocol. Purified cells (100,000 cells per well; purity >95%) were plated in 96-well plates with complete RPMI medium.

Human CD14⁺ cells and mouse BMDMs or BM monocytes were infected with M. tuberculosis HN878 isolate at a multiplicity of infection (MOI) of 2 (2 bacteria:1 cell). Extracellular bacteria were washed away with PBS after 4 h and bacterial growth was quantified by CFU enumeration 4 h (control) or 4 days after infection. The culture supernatants (BMDMs and BM monocytes) were collected 24 h (in the absence of bacterial washing) after infection for protein quantification.

Cytokine production by BMDM or monocyte cultures infected with M. tuberculosis was quantified by immunoassay, using commercially available ELISA kits (Cat. #88-7064-88; 88-7013-88; 88-7324-88, Invitrogen).

Data were plotted and analyzed using GraphPad Prism (version 8.1.0): a non-parametric Mann–Whitney test was used to compare two groups and Kruskal–Wallis one-way analysis of variance test was used for the comparison of more than two groups. Multiple comparison correction was performed using a Dunn post-test. Significant differences are as follows: * p ≤ 0.05; ** p ≤ 0.01; and *** p ≤ 0.001. Spearman correlations and correspondent correlation plots were obtained using R (version 4.1.2) and R package corrplot (version 0.92).

To investigate possible alterations imposed by M. tuberculosis infection to the BM hematopoietic output, we aerosol-infected WT C57BL/6 mice with high doses of M. tuberculosis HN878. Twenty-seven days post-infection (p.i.), the mice presented a high bacterial burden in the lungs (over 10⁹ bacteria) ( Figure 1A ), which was accompanied by marked tissue pathology, with lesions affecting 14% to 56% of the lung ( Figure 1B ). Additionally, we observed an increase in lung cellularity, associated with the recruitment and accumulation of CD45⁺ immune cells ( Figure 1C ). We next analyzed by flow cytometry the myeloid cell composition within the immune cells present in the lung, which accounted for approximately 45% of the total CD45⁺ cells ( Figure 1D ; Supplementary Figure 1A ). Although the frequency of alveolar macrophages decreased with infection, resulting in a very low percentage 27 days p.i., both the frequency and numbers of all other myeloid cell types (monocytes, monocyte-derived recruited macrophages, dendritic cells (DCs), and neutrophils) were significantly increased in infected mice ( Figure 1D ). Neutrophils exhibited the highest increase both in percentage and cell counts of the myeloid populations analyzed. The global frequency of other immune cells (which included eosinophils and NK, T, and B cells) was not altered ( Figure 1D ). Immunofluorescence staining of lung sections with Mpo, a neutrophil marker, revealed large Mpo-positive areas, which were dominant compared with those corresponding to Nos2-positive activated macrophages ( Figure 1E ). These results are in line with previous reports showing an accumulation of myeloid cells, especially neutrophils, in the lungs upon infection with M. tuberculosis HN878 (28, 31).

We then analyzed the blood and BM compartments at 27 days p.i. ( Supplementary Figure 1B ). Infected mice presented leukocytosis ( Supplementary Figure 1C ), which was associated with increased numbers (per µl of blood) of monocytes, neutrophils, and T cells, even though the proportion of T cells in the blood had diminished ( Figure 1F ). There was a reduction in the frequency of B cells, with no effect on cell numbers ( Figure 1F ). Thus, the increase in myeloid cells seen in the lungs of infected mice was also observed in the blood, suggesting a possible altered hematopoietic output. Analysis of the BM mature cell populations showed an increase in the frequency and numbers of monocytes, which was accompanied by a decrease in the frequencies and numbers of B cells ( Figure 1G ). Both neutrophils and T cells remained mostly unaltered ( Figure 1G ), as did the total number of BM cells ( Supplementary Figure 1D ). Importantly, M. tuberculosis disseminated into the BM ( Figure 1H ). Targeted transcriptional analysis of the BM 27 days p.i. showed a modest impact of infection on the expression of cytokine genes involved in hematopoietic reprogramming. Tnf and Csf1 expression was statistically significantly increased and Csf2 expression was statistically significantly decreased ( Figure 1I ). The lack of more profound differences in gene expression and the decrease of Csf2 transcription upon infection was surprising and may reflect the overall exhaustion of the infected BM.

The accumulation of myeloid cells in the lungs of the infected animals, together with the alterations seen in the blood and BM, raise the hypothesis that EM may be happening in this immune compartment.

To investigate whether EM may be occurring during M. tuberculosis HN878 infections, we analyzed the BM progenitor populations of naive versus HN878-infected mice 27 days p.i. ( Supplementary Figure 2A ). UMAP dimensionality reduction analysis of flow cytometry data revealed a different precursor and progenitor cell composition profile between infected and non-infected mice ( Figure 2A ). Closer analysis of the different precursor populations revealed a statistically significant increase of LSK (Lin^-Sca⁺c-Kit⁺) cells in infected mice ( Figure 2B ). Within the LSK population, long-term HSC (LT-HSC) were not affected by the infection, but an increase of the multipotent progenitors (MPP) 2, 3, and 4 was observed ( Figure 2B ). Moreover, although no statistically significant differences were observed for the frequencies of LK (Lin^-Sca^-c-Kit⁺) cells ( Figure 2C ), there was a statistically significant increase in granulocyte-monocyte progenitors (GMPs) accompanied by a decrease in common myeloid progenitors (CMPs) and megakaryocyte erythrocyte progenitors (MEPs) ( Figure 2C ). Similar alterations were seen in the frequencies of precursor and progenitor populations ( Supplementary Figures 2B, C ). Altogether, these data suggest the occurrence of EM as a consequence of aerosol infection with high doses of M. tuberculosis HN878.

Given the apparent discrepancy between these data and previously published data (24), we decided to infect mice with low doses of M. tuberculosis HN878. Under these conditions and at day 34 p.i., we still observed colonization of the BM ( Supplementary Figure 2D ), although the bacterial burden was 2 log lower than that resulting from infections with high doses of bacteria. Aerosol infection with low doses of M. tuberculosis HN878 did not significantly remodel BM cell composition, as no changes in mature cell populations were observed ( Supplementary Figure 2E ), and only an increase in LSK ( Supplementary Figure 2F ) and a modest decrease in CMPs ( Supplementary Figure 2G ) were observed. Interestingly, a positive correlation was observed between the BM LSK population and lung bacterial burden, monocytes, and neutrophils, as well as BM bacterial burden ( Supplementary Figure 2H ).

Collectively, our findings raise the hypothesis that EM may occur during M. tuberculosis infections, possibly associated with more severe outcomes of disease, such as those resulting from high doses of infection.

As we observed the adaptation of the BM towards myelopoiesis in our model of aerosol infection with high doses of M. tuberculosis HN878, we next investigated whether the resulting mature cells might be trained. As such, we generated macrophages derived from the BM of non-infected or M. tuberculosis-infected mice for 27 days. The resulting BMDMs were infected with M. tuberculosis HN878 and the bacterial burden was measured 4 h and 4 days p.i. The bacterial burden 4 h after infection was similarly independent of the BMDM origin. Then, an increase in intracellular bacteria was observed over time and was similarly independent of the BMDM origin ( Figure 2D ), suggesting that previous infection with M. tuberculosis did not enhance the protective capacity of de novo-generated BMDM. Additionally, we measured the production of TNFα, IL-6, and IL-1β, cytokines associated with trained immunity in the context of mycobacteria (35, 36), by BMDMs generated from non-infected or infected mice and in vitro infected with M. tuberculosis HN878. The ability of the generated BMDMs to respond to secondary stimuli was also not enhanced by previous infection across the measured parameters ( Figure 2E ). Finally, to exclude possible effects imposed on macrophages by the in vitro differentiation process, we purified monocytes from the BM of non-infected or M. tuberculosis-infected mice and analyzed their microbicidal capacity ( Figure 2F ) and cytokine response to secondary stimuli ( Figure 2G ). Monocytes purified from non-infected or infected mice responded similarly, further supporting that despite evidence of EM in the BM, this was not accompanied by training of the mature populations.

As trained immunity was suggested to equip circulating monocytes from individuals recently exposed to M. tuberculosis with an increased capacity to control BCG (37), we considered the hypothesis that trained immunity may occur at the periphery. To test this hypothesis, we started by comparing the bacterial growth over a 4-day period in monocytes isolated from IGRA ⁻ or active TB patients. In IGRA ⁻ monocytes the bacteria were able to grow; however, this growth was much more limited in the case of active TB monocytes ( Figure 2H ). Moreover, when we compared the ability of monocytes from active TB patients to control M. tuberculosis before and after antibiotherapy, we found that after treatment the growth of bacteria was superior for all tested donors ( Figure 2I ).

Altogether, our data demonstrate that in a mouse model mimicking the human transcriptional signature of TB, BM monocytes or macrophages did not acquire trained immunity despite the occurrence of EM. However, untreated active TB in humans appeared to associate with circulating monocytes that display an increased capacity to control M. tuberculosis growth in in vitro infections.

As IFNγ has been widely implicated in EM (38, 39) and we observed an increase of the expression of this cytokine in the BM of M. tuberculosisHN878-infected mice (Figure 1I), we sought to investigate whether its presence was required for the EM observed in the C57BL/6 model of infection. Thus, we aerosol-infected C57BL/6 WT and IFNγ^-/- mice with a high dose of M. tuberculosis HN878. IFNγ^-/- mice presented higher lung bacterial burdens than WT mice, in agreement with their known inability to cope with M. tuberculosis infections ( Figure 3A ). In the same line, histological analysis revealed more extensive lesions in IFNγ^-/- mice than in WT mice, with damaged tissue and pulmonary exudate visible in areas adjacent to lesions ( Figure 3B ). Both the percentage of lesion ( Figure 3C ) and the lesion score ( Figure 3D ) were increased in infected IFNγ^-/- mice compared with WT mice. A lack of IFNγ did not impact the number of total or immune (CD45⁺) lung cells ( Figure 3E ). The numbers of monocytes, recruited macrophages, and neutrophils present in the lungs of IFNγ competent or deficient mice were also similar ( Figure 3F ). Despite this overall similarity in myeloid cell composition in the lungs of infected WT and IFNγ^-/- mice, the increase of monocytes and neutrophils seen in the blood of infected WT mice was not observed in the absence of IFNγ ( Figure 3G ). We then focused on analyzing the BM compartment. As observed in the lung, IFNγ^-/- mice presented a higher bacterial burden in the BM than WT mice ( Figure 3H ), with BM cellularity ( Figure 3I ) and the variations in the number of monocytes ( Figure 3J ) similar in the two mouse strains. However, IFNγ^-/- mice exhibited a reduction in BM neutrophils compared with WT mice ( Figure 3J ). These data indicate that, in the absence of IFNγ, the accumulation of monocytes and neutrophils observed in the lung is not fully reflected in the blood and the BM, as seen for WT mice. Although this finding is not fully understood, it may suggest a faster recruitment of neutrophils to the lung, with increased exit rates from the BM, or alternatively an increased lifespan of neutrophils in the lung.

Next, we questioned whether EM would occur in IFNγ^-/- mice upon M. tuberculosis HN878 infection. Similar to the WT controls, UMAP reduction analysis of flow cytometry data disclosed differences in precursor and progenitor cell composition profiles between infected and non-infected IFNγ^-/- mice ( Figure 3K ). The analysis of precursor populations revealed an overall similar tendency between WT and IFNγ^-/- mice upon infection, with an increase of all populations in infected mice compared with uninfected mice ( Figure 3L ). However, for some populations, such as LSK, MPP2, and MPP3, this increase was less marked in the case of IFNγ^-/- mice, ( Figure 3L ). A similar profile was observed in progenitor populations, with less pronounced alterations seen in IFNγ^-/- mice ( Figure 3M ). Of note, the most marked change observed in IFNγ^-/- mice upon infection was an increase in the GMP population ( Figure 3M ). Similar findings were observed in terms of cell population frequencies ( Supplementary Figure 3 ). Altogether, these findings suggest that EM in response to severe M. tuberculosis infections might occur, at least to some extent, even in the absence of IFNγ. EM has been linked to inflammatory environments in the BM (40). Thus, we questioned whether the lack of IFNγ would alter the inflammatory environment seen in the BM of infected WT mice ( Figure 1I ). Interestingly, the transcription of several cytokine genes in the BM was decreased in infected IFNγ^-/- mice compared with WT or uninfected mice ( Figure 3N ). However, this was not the case for TNF ( Figure 3N ), the transcription of which was enhanced by the infection independently of IFNγ ( Figure 3N ).

Together these data suggest that IFNγ is partly dispensable for the occurrence of EM upon the infection of mice with M. tuberculosis HN878, consequently suggesting the occurrence of alternative mechanisms.